curious independent builder
i build things to make everyday life a little less complicated. mostly software, automation, and small tools that solve problems i run into.
i care about useful things, simple solutions, and taking the time to make them work well.
a few things i’m working on at the moment.
making scheduled recovery more reliable and administration easier to use safely.
improving how kontrole & vilciens handle changing information while respecting privacy.
helping supervised android tasks recover from interruptions and stay quiet when idle.
keeping my services within their resource limits and preparing projects for public release.
ongoing work; the projects below describe what’s already in place.
ongoing systems, useful little tools, and a few things i learned from.
systems i’m building and looking after.
browser and phone
a browser interface for requesting supervised actions on an android phone and seeing the result.
access, phone readiness, or a dropped connection can interrupt a simple task.
access controls, supervised phone actions, health checks, reliable scheduling, and clear results.
the project is linked below. device identities, access details, control codes, ticket data, and private state are kept out of this page.
public transport
maps, transport updates, and live train views for getting around riga and beyond.
useful travel information is scattered across stops, routes, and changing reports.
web and telegram tools for maps, incidents, departures, and train movement, with guides in multiple languages.
the public sites show how it works. private messages and internal analysis stay private.
phone automation
a reusable runtime for supervising android tasks and recovering from interruptions.
phone tasks depend on app state, device state, long-running processes, and unreliable connections.
a kotlin and adb runtime with task supervision, health reporting, and recovery for ticket remote and other phone workflows.
the public pixel-phone repository documents the runtime. device identifiers, private endpoints, account state, and logs stay private.
running services
the setup i use to deploy and run go services on my own infrastructure.
even small services need repeatable releases, resource limits, visible failures, and a way back when something breaks.
a linux and docker setup for releases, monitoring, health checks, bounded diagnostics, dns management, and rollbacks.
public-ops shows the structure and operating practices. hosts, addresses, ports, credentials, dashboards, and private state are omitted.
tools for publishing work and managing access.
publishing tools
a way to turn local projects into readable public repositories and keep them in sync.
working folders often mix useful code with private details, generated files, and missing context.
a python package for finding, reviewing, cleaning, summarizing, publishing, and syncing projects, with an explicit choice about what becomes public.
the main repository documents setup and checks. its frontend and backend mirrors are parts of the same project.
private project
a membership mini app for access, renewals, and a clear view of account status.
access, renewal, payment-related state, and support need to make sense together.
a private access and renewal workflow with a mini app interface and clear status for each user.
only the broad design and reliability approach are described here. business, payment, customer, account, and recovery details remain private; there is no public repository.
a particular problem, a focused solution.
maps and data
a searchable map that turns spreadsheet measurements into something easier to explore.
location and connection details are hard to compare row by row. raw exports can also contain private columns.
linked map points and records, summaries, search, and filters. ip-address fields are excluded.
explore the published speed measurements on the map. ip-address fields stay out of the public dataset.
interactive web
a quiet particle scene that opens into an interactive game.
real-time graphics need to work with modest hardware, touch, keyboards, contrast, and reduced motion.
an adaptive simulation with ambient and game modes, multiple inputs, saved state, haptics, and performance controls. webgpu has a canvas 2d fallback.
both pages are public. checks cover rendering, timing, contrast, sessions, and input.
macos utility
a native app for viewing and switching saved codex sign-in contexts.
similar-looking account files make it too easy to switch to the wrong one.
a swift app that finds saved contexts, shows the active one, and makes switching deliberate, with a command-line fallback.
the public repository explains the flow. credentials, tokens, and active account files are excluded.
data tools
a parser that gathers latvian court calendars into one searchable workbook.
hearings span monthly pages, different case types, and an older text encoding.
a crawler that reads windows-1257 pages, aligns shared and case-specific fields, adds cited law text and summaries to criminal cases, and exports to excel.
the repository shows the parsing steps and output structure, so the transformation can be checked.
scheduled tasks
a small public layer for scheduling maintenance while keeping application work private.
job schedules should be easy to review without exposing sensitive endpoints or business operations.
shell scripts, github actions, and documentation that show when jobs run and how they start.
the repository exposes the scheduling setup. the application-specific operations it calls remain private.
a small policy file
a versioned dns blocklist with a clear place to review changes.
a rule hidden in a local file is hard to find, discuss, or maintain.
a public policy definition with context and a tracked history.
the policy and its history are readable directly on github. a document is enough for this job.
file utility
a small python script that fixes duplicate json filenames in google takeout exports.
duplicate numbers can end up after the wrong extension, making a large export awkward to clean by hand.
a command-line script that separates the name, duplicate number, and extension, then puts them in a consistent order.
the repository includes the script, before-and-after examples, and an overwrite warning to review before running it.
earlier tools, experiments, and forks, with credit to their original projects.
personal automation
a telegram-controlled watcher that sends a message when unread counts meaningfully change.
repeated inbox checks waste attention, while noisy notifications just create another task.
a python service with start, stop, status, debug, check-now, sign-in, and help commands, adaptive polling, session upkeep, retries, and health checks.
the repository explains the controls, quiet-until-changed rule, container setup, and handling of secrets and expired sessions.
fork and experiment
an experiment in controlling codex from a phone and macos through a cloudflare tunnel.
remote access needs clear boundaries around the host, bridge, client, and tunnel.
a macos host, rust bridge, expo client, policies, health checks, tunnel setup, and a legacy typescript path. the bridge stays on localhost before controlled exposure.
this is a fork of an existing project. the public repository credits the upstream work and documents the components and setup.
archive
two automation forks kept as a record of things i explored.
older experiments deserve context and clear credit to the people who started them.
experiments with the duolingo and microsoft e5 auto-renew projects. these are forks, not maintained original releases.
their fork relationships remain visible on github. they are included as learning history.
a confusing journey, a repetitive check, a messy export. i like finding the small problem that keeps taking up attention.
sometimes a page helps. sometimes a script or a document is enough. i try to choose the smallest thing that makes a difference.
i check what people actually see and use: the page, the phone, the links, and what happens when something goes wrong. private data stays private.